創新擴散理論
(Innovation Diffusion Theory ,IDT) &
Information technology innovation diffusion:
an information requirements paradigm
心得報告
指導教授：許芳銘 教授
學
生：陳俊勳
中華民國 一零二 年 三 月 十六 日
1
摘要
許多資訊科技都是創新事物，要在社會或組織中被廣
泛使用，需要有擴散的過程。因此有些研究便在探討這個
變化的過程及其影響的因素。其中最具代表性的理論是創
新擴散理論(Innovation Diffusion Theory,IDT) 。本章
介紹創新擴散理論的基本概念及它在資管領域的應用。主
要內容如下:
 創新擴散的定點
 創新擴散理論的內涵
 創新擴散的模型
 創新擴散的應用
2
導論



創新擴散理論(Innovation Diffusion Theory,IDT)是研究創新事物
在社會體系間，在一定時間內，透過一定管道所傳播的過程。創新
擴散理論指出，任何一種新事物(如新觀念、新發明、新風尚、新科
技、新產品等)，從誕生到逐步被社會大眾所接受而流行起來，都會
經歷一個在社會體系中推廣或擴散的過程。因此創新擴散的模式成
為一個重要的研究課題。在資訊管理的領域中，系統(如ERP)導入組
織而被採用的過程，是創新擴散理論的應用之一。資訊委外概念的
擴散也曾是很有趣的應用。
創新擴散理論的研究最早從1903年開始，並持續至今。該理論自學
者Rogers(1995)提出創新擴散模型後，更廣泛的被應用在行銷管理
、人類學、社會學、教育研究、公共健康與醫藥社會學、經濟學等
領域。
在本章中，我們在第二節說明理論的起源與「創新」、「擴散」的
定義，第三節則解釋影響創新擴散的四大因素與創新採用者的類型
，第四節以Bass(1969)的新產品擴散模型說明創新擴散的數學模型
，最後則以Hu et al.(1997)的論文為例。說明創新擴散模型的資管
領域的應用。
3
創新擴散理論的起源


有關創新擴散理論的研究最早可以推溯至1903年，是由Gabriel
Trade對社會科學的研究開始，之後有不少學者分別提出自己的看
法或應用。例如「首次購買擴散模型」是1960年由Fourt &
Woodlock研究食品雜貨產品所提出的模型，1961年Mansfield研究
產業創新之科技替代，並提出其模型架構。
但創新擴散理論模型之集大成者，則是美國新墨西哥大學的傳播與
新開學教授Everelt Rogers，他在1995年對超過3,000個創新擴散
的案例加以彙整，總結描述創新事物在一個社會體系中擴散的基本
規律和過程，因此Everelt Rogers也被稱為「創新擴散理論之父」
。
4
創新的定義



「創新」(Innovation)可以是一種概念(Idea)、實務(Practice)或是
具體的事物(Object)，或任何可被人或組織認為是一種新的事物
(Rogers,1995)。而創新一詞具有多重面向，可以是一個過程
(Process)，同時也可能是一個結果(Outcome)；可以是一種內在的反
應，亦可能是外在的改變(Damanpour,1996)。Rogers(1995)認為「創
新」(Innovation)等同於「科技」(Technology)，而科技是「為了達
成某些預定目標，以具體行體降低其因果關係中的不確定性的一種設
計」。
資訊科技的創新包含硬體與軟體兩個構成元素(Components)，而管理
創新則包括概念與流程：
 硬體(Hardware)：指科技實體的部分，由材料或實際物體所構成
之物件。如電腦是由半導體、電晶體、積體電路等硬體元件所組
成的。
 軟體(Software)：指以資訊為基礎的工具。電腦軟體包括作業系
統、系統程式等，可以協助人們解決特定問題的工具。
通常我們提及科技創新時，多是指硬體的部分，但實際上，科技本身
是軟體的混合，單只有硬體(如電腦主機)，並無法充分發揮作用，而
必須配合適當的軟體，才能發揮實質效益。
5
擴散的定義



Rogers(1995)界定「科技創新」(Technological Innovation)包含以下
兩項：
 軟體資訊(Software Information)：包含在科技中，可以降低在因
果關係推導中的不確定性，進而達成預期效果。
 創新評估資訊(Innovation-evaluation Information)：一個創新要
成功最重要的是要有一個評估衡量創新績效之機制，作為改善創新
系統及制定創新政策之參考，這些資訊可以降低創新本身的不確定
。
擴散是個過程，即在一定時間內，社會體系中的成員藉由特定管道，傳
遞某項創新訊息的過程。它是一種特殊形式的溝通，其溝通的訊息與創
新事物有關，而溝通雙方或多方就此特定議題交換意見，尋求共識
(Rogers,1995)。所以擴散(Diffusion)是一種社會化的過程
(Merton,1973)，某個特定的構想或產品，在某個社會體系中，透過溝
通管道，經過一段時間後，被社會體系的成員所接受(Katz et
al.,1963)。
擴散過程中資訊與通路扮演重要的角色。由於擴散過程是在傳遞新的構
想或事物，而通常人們對於新的構想或事物總是感到陌生，難免會產生
程度不一的不確定性(Uncertainty)，即事物的不可預知，結果不全和
訊息不足，而充足的資訊可以降低這種不確定性，以利資訊接收者解決
問題。
6
創新擴散的內涵


Rogers(1995)對創新擴散的定義是指創新事物透過某些傳播管道，經
過一段時間後，被社會體系採用的過程，因此在創新擴散的過程中有
四個主要的元素：
「創新」(Innovation)、「時間」(Time)、「傳播管道」
(Communication Channels)及「社會關系」(Social System)。
社會系統
時間
創新
創新擴散
被接受速度
傳播管道
圖4-1 創新擴散過程的活動架構
7
影響創新擴散的元素-創新(Innovation)







(一)創新(Innovation)：影響創新是否被接受的第一個因素是創新事物的特質。個人或
組織對創新事物特質的認知，會決定他們是否採用創新事物，如果創新事物比起舊事物
具有相對優勢，能夠為個人或組織帶來較大的利益或滿足感等，創新事物較容易被接納
。通常創新事物具備下列幾項特質時，較容易被接納：
1.相對優勢(Relative Advantage)：創新產品的功能被認為比舊產品更優秀(例如，重
量較輕的筆記型電腦就比大而重的具有輕便的慢勢)。決策者對「創新」所能認知的相
對優勢愈多，則採用的速度就愈快，亦即擴散速度會愈快。
2.相容性(Compatibility)：創新產品與個人原有的價值觀，業界標準、組織架構或過
去的經驗，及目前的需求的契合程度。使用者愈不需要改變自己來配合新產品，便較有
可能提早採用。例如，新款的2G手機因為和舊系統相容，因而很容易銷售，但3G手機剛
出來時，因與當時的2G標準相容性低，初期採用者便很少。
3.複雜性(Complexity)：創新產品若難以了解或使用，則採用會較慢，一項新產品若使
人覺得難以了解其品質，效益或或者操作過於複雜，就會延遲被採用的時間。
4.試用性(Trialability)：創新產品可被試用的程度也會影響採用的意願。當一項新
產品可以讓消費者僅花費極少的力氣，會有利於人們採用此項新產品。
5.可觀察性(Observability)：創新產品的效益或功能愈容易被理解或說明時，會使越
多的潛在採用者(Potential Adopters)很快明瞭，則該產品的相關資訊流通速度也會加
快。則會促使更多人盡早採用。例如：iPad輕巧的設計使消費者很容易看到其價值。
過去研究顥示在影響創新擴散的原因中，創新事物的特質具有49%到87%的解釋能力
(Volink et al., 2002)，說明創新本身對擴散的技術有較高相容性、具有可試用性，
可觀察性及較低的複雜度，則此創新事物將有機會很快的擴散到社會中。
8
影響創新擴散的元素-時間(Time)







(二)時間(Time)： 創新擴散是一個過程，它指的是經時間積累，社會體系的成
員通過選擇某一管道，對某一創新發明的採納。由於這種決策過程是個別成員的
自主決定，而不足權威性或者是集體性的，每個成員的採納過程大致要通過以下
五個步驟：
1.「知識」(Knowledge)階段：知道這項創新的存在，而且對它的功能有了初步
的瞭解(例如，知道3G手機可以更方便的下載多媒體內容)。
2.「說服」(Persuasion)階段：對創新事物形成認同或不認同的態度(例如，認
為3G手機有沒有需要)。
3.「決策」(Decision)階段：採取行動做出採用或拒絕這項創新的決定(例如，
認為有必要，並決定去買一支3G手機來使用)。
4.「實踐」(Implementation)階段：開始使用創新事物(例如，到通訊行買了新
的手機)。
5.「確認」(Confirmation)階段：使用之後有更清楚的瞭解，使用者會繼續使用
或效果，並將個人經驗傳播給其他的潛在使用者，可能會影響他們採用創新事物
的態度。
前述個人對創新事物的決策流程，可以用圖4-2加以顯示，通常創新決策過程的
五個步驟是依照時間順序出現的，每個人的創新決策期都不同，有的人從獲得創
新資訊，到確切採納的過程十分快速，但有的心必須花好幾年的時間才能接受一
項創新(不同階段的採用者可參考圖4-3)。
9
過程
1知識
接受者特質
1. 人口所得特徵
2. 個性
3. 溝通行為
4. 個人需求
2.說服
創新的特質
相對優點
相容性
複雜性
可試用性
可觀察性
結果
溝通管道
3.決策
採用
拒用
4.實踐
5.確認
繼續採用
較晚採用
不再採用
繼續拒用
10
影響創新擴散的元素-傳播管道(Communication
Channels)



(三)傳播管道(Communication Channels)：Rogers(1995)強調無論是哪一種新的
發現和創新，相關資訊都必須經過適當的傳播管道來流通。在擴散的過程中，能
否吸引人們的注意，便成為此一創新事物最後能否被成功推廣的關鍵因素。
Rogers(1995)將創新資訊的傳播管道分為「大眾傳播媒體」(Mass Media
Channels)及「人際的溝通」(Interpersonal Communication)兩種。大眾傳播媒
體是讓潛在採用者最快、最有效率獲得創新事物存在的方法，如廣播、電視、報
紙等。而個人間的溝通管道則是說服他人採用創新事物最有效的方法，特別是溝
通雙方(或多方)具有同質性(homogeneous)時，效果會更佳。同質性意指溝通雙
方在社會經濟地位，教育程度或宗教信仰等方面有相似的背景。一般而言，大眾
媒介在知識階段比較重要，而人際溝通則在說服階段比較重要。
在資訊領域，我們可以把資訊傳播管道分為「非正式傳播」、「正式傳播」、及
「電子化傳播」三種形式。「非正式傳播管道」(Informal Communication
Channel)是指存在於團體之間，由個人來進行完成的各種資訊交流及人際的溝通
；「正式傳播管道」(Formal Communication Channel)指的則是透過正式的出版
品或發行的資料，或大眾傳播媒體等進行的資訊傳播。至於「電子化傳播管道」
Electronic Communication Channel)的利用，則是在電子科技昌盛的今日，許
多舊有的傳播管道都逐漸被這種新世代的科技所取代。
電子化傳播管道的特色就在於它是個人溝通媒介(非正式傳播管道)與大眾傳播媒
體(正式傳播管道)的結合，而且這個傳播管道還是雙向的。換言之，個人不只是
傳統的單向訊息接收者，只要他願意，也隨時可以主動地搜尋他所需要的資訊，
甚至也能夠扮演資訊的製造者與傳播者。例如近年風行的部落格或網站社群媒體11
，就是一個電子線上傳播管道最具有影響力的機制。
影響創新擴散的元素-社會體系(Social







System)
(四)社會體系(Social System)：社會體系是指一群相互關聯的單位，一起參與解決共同面
對的問題，並達成共同的目標。單位可能包括個人、非正式團體、公司組織等。社會體系
中的社會結構(Social Structure)是影響創新擴散的最主要因素。社會結構分為正式結構
(Formal Structure)與非正式結構(Informal Structure)，前者是指如政府部門般業務分
工、上下權責分明的組織；後者則是人與人在特定情境下的互動，例如志趣相同或有共同
需要的人成為朋友。在社會結構中的成員，如果是同質性，就容易形成一個溝通結構
(Communication Structure)，彼此交換資訊。
但創新事物要進入一個社會體系中被接受與擴散，則需要成員間具備若干的「異質性」
(homogeneous)。因為溝通雙方同質性高，則彼此擁有的創新資訊一樣，自然無需交換，創
新也比較無法擴散。最好的狀況是溝通雙方在創新的看法上是異質性的，但在背景上有某
種程度的同質性，才能在互動的過程中交換創新的資訊，有利於創新的擴散。
因此，具有一定程度的異質性社會，人們對於創新的接受程度較大，社會整體也傾向鼓勵
變化，對創新的啟發與激勵具有正面推力。而同質性的社會體系傾向抗拒變化與創新，因
為團體間的凝聚力強，團體決策壓力也就容易出現，異見容易被壓仰，在這樣的社會體系
中，創新事物的擴散比較緩慢，甚至有些產品無法擴散。
在創新擴散的過程中，有三種重要的角色存在：意見領袖、變動推手及變動助手。
意見領袖(Opinion Leaders)：對其他人有很大影響，他們的看法往往會帶動別人跟進。例
如，有些科技領袖極力提倡行動商務或某些科技便會帶動風潮。
變動推手(Change Agents)：協助創新被引進社會系統中，被大家接受。例如，政府機構會
編列預算推動某些新科技。
變動助手(Change Aids)：協肋變動推手與潛在使用者密切接觸的人。例如，有些單位會輔
12
導科技的導入或協助解決導入的問題。
創新採用者的類型




影響創新被採用的速度，採用者的特性當然也是個重要因素，在行銷領域中，創
新的採用者被分為五種類型，每一類採用者皆有較為顯著的行為模型與作用。這
五類使用者的分佈與比例接近常態分配。
創新者(Innovators)：具冒險精神，對於新事物充滿好奇，不計較物品價格，在
創新商品剛推出的時候就迫不及待的想購買。通常創新使用者本身都有非用不可
的理由，例如台灣第一批使用手機的人是業務人員及公司老闆，為了要隨時都能
方便地接到電話，成為早期手機的創新採用者。另外，有一些人是熱愛創新事物
的人，會購買的原因純粹是因為對創新事物的嘗鮮興趣，而且這些人也會形成小
團體，彼此交流心得，在網路上討論，有新的商品出來了會爭相走告，而且通常
會有自己捍衞的品牌，忠誠氷十分高。Rogers認為創新者所佔的比例不多，約只
有2.5%，早期他更認為這些創新者只分佈在最初購買的期間。
早期採用者(Early Adopters)：早期採用者是最早受到創新者影響的人。例如看
到家人用手機，或看到同事用PDA似乎很好用，就會受影響而自己買一台來用。購
買之前，他們會詢問創新者的意見，作為採購參考。這類早期採用者約佔全部市
場的13.5%。創新者影響早期採用者的管道，通常是透過人際網絡而引起興趣。
早期跟隨者(Early Majority)：當新事物已有前兩類使用者採用後，接著會有佔
了34%的早期跟隨者，加入使用的行例，但是他們與前兩者有不同的需求，因為市
場上已有許多人採用，也對該創新有一定的資訊，因此早期跟隨者們最希望的是
能「一指搞定」，比較容易被簡單的訴求打動。例如大陸個人電腦的第一品牌聯
想電腦，在銷售電腦這種高科技產品時的訴求只有一個，「別讓自己的寶貝孩子
輸給別人」。行動電話在台灣的普及只花了三年的時間。而各家業者最早廣告訴
求居然只有簡單的「打得通，接得到」。
13
創新採用者的類型



晚期跟隨者(Late Majority)：接下來會採用的一批人稱為晚期跟隨者，也佔整
體的34%，是比較後知後覺的人。他們往往在產品進作成熟期，社會上大部份的
人都已經在使用某種產品了，才開始想要使用。購買該產品的目的有時候不是為
了產品本身，反而是來自社會的壓力。「別人有的，我也有就好」常常是這種使
用者的心態，而且希望產品「買最便宜的就好」，因為此時創新事物往往已經進
入成熟市場的價格戰當中。例如，電視機的市場幾乎已經飽和而且利潤固定了，
這時強調產品「價格一樣，功能卻最多」可能就是有效的訴求了。
落後使用者(Laggards)：最後仍然會有少數人對採用新科技不感興趣，這些約佔
了16%的落後使用者，往往是傳統保守的採用者。他們可能因為自身對新產品的
疑慮或認為沒有需求，而抗拒採用新產品。他們可能非常傳統，或與社會體系隔
離，可能對創新發明表示懷疑，並經常的與那些有著傳統價值觀的人交流互動，
因而拒絕採用。例如，有人因為不喜歡受干擾，因此至今仍不用手機。
根據Beal & Rogers(1960)的統計，上述五種採用者，在採用創新的決策時間上
極為不同。創新者僅需0.4年即會採納創新事物，但落後跟隨者則需要長達4.65
年，要花將近11.5倍的時間才會接受創新。
14
創新擴散模式
GAP
採
用
者
類
型
與
比
例
創新者
2.5%
早期大眾
晚期大眾
34%
34%
早期採用者
13.5%
x  2
x 
x
落後者
16%
x
創新接受的時間
時間
創新擴散的S形模型

創新擴散模型的研究最早有Fourt & Woodlock(1960)，Rogers(1976)、
Haines(1966)等學者針對產品創新、創新擴散與採用、耐久財的擴散等方向進行
研究，但在新產品創新擴散的過程中，由於影響因素很多，因此出現以量化模型
進行新產品擴散過程的分析，其中以Bass(1969)的創新擴散模型最具代表性。以
下便以Rogers/(1976)與Bass(1969)的創新擴散模型來說明。
16
Rogers的創新擴散模型


Rogers(1976)認為創新事物在特定社會體系中的擴散，是呈現S形曲線的擴散型
態(如圖4-5)。他認為一種創新在一個社會體系中的擴散，當使用者達到體系中
總人口的某一比例後，整個擴散過程就可以自續(Self-Sustaining)下去，這一
比例(有時也用絕對數量，即已採用者的數量)，就是臨界數量。
通常當一種創新剛剛開始在系統中擴散時，人們的接受程度比較低，因此開始的
擴散過程比較緩慢；而當採用者比例一旦達到臨界數量後，擴散過程就會加快，
出現起飛(Take-Off)的現象。當系統中大部分最終會採用創新的人都在這一階段
採用該創新之後，擴散過程會再次慢下來，系統對創新的採用逐漸達到飽和點
(Saturation Point)。以上的過程使創新擴散的速率成為常態分佈曲線(如圖45)。
17
創新採用者的決策時間
(Beal & Rogers, 1960)
創新者
早期採用者
早期跟隨者
晚期跟隨者
落後使用者
0
1
2
3
4
5
(…
18
Bass的新產品擴散模型

Bass(1969)的新產品擴散模型則涵蓋了Fourt & Woodlock(1960)與Mansfiled
(1961)的理論差異，重點在訊息傳播的途徑。前者認為在擴散的過程中，新產品
的潛在消費者只受大眾傳播媒體的影響；而後者則認為潛在消費者是受到採用者
以口碑方式行銷的非正式管道影響。Bass結合了兩者的影響，因此潛在採用者可
分為受到大眾媒體影響(外部影響，External Influence)的「創新者」，及受到
採用者口碑影響(內部影響，Internal Influence)的「模仿者」。此外Bass模型
也隱含一個基本假設：「在特定期間內，每位顧客的購買量均為一個單位，而且
沒有重複購買的情形出現」。
19
Bass的新產品擴散模型


基於上述的假設，Bass的創新模型(如圖4-6)有三大特性：
(1)累積採用者的分配圖形呈現S型的曲線。
(2)當期的新增購買者數量，具有鐘型(Bell)的常態分佈曲線型態。
(3)鐘型曲線是對稱的(Symmetric)
Bass模型源自於「危險函數」(Hazard Function)。危險函數為新產品在t時之前未
曾被採用而在t時會被採用的機率，其計算公式為：
prob =

基於危險函數的公式，Bass計算創新擴散的公式如下：
𝑓(𝑡)
1−𝐹(𝑡)

𝑓(𝑡)
1−𝐹 𝑡
= p + qF(t)
其中，f(t)為消費者在第r期採用該產品的比率。
F(t)為f(t)的累積機率分配，即累積已採用該創新的人數比率。
p 為創新係數，即創新者因多媒體或其他資訊管道影響而採用創新的效果，即外
部效果。
q 為模仿係數，即尚未採用者中會因見到許多已採用者F(t)而加以採用的比率，
為內部效果。
20
Bass的新產品擴散模型

若m為最終採用創新物的總潛在人數(採用者的人數上限)，則t期的採用人數為：
𝑡

N(t) = mF(t) = m

n(t) = mf(t) = m

𝑓(𝑡)
1−𝐹(𝑡)



1−𝑒 −(𝑝+𝑞)
𝑞
𝑝
𝑡
1+ 𝑒 −(𝑝+𝑞)
𝑝(𝑝+𝑞)2 𝑒 −(𝑝+𝑞)
(𝑝+𝑞𝑒
𝑡
−(𝑝+𝑞)𝑡 2
)
= p + qF(t) = p + q
𝑁(𝑡)
𝑚
其中，n(t)：為t時的當期採用人數。
N(t)：為t時的累積採用人數。
若對N(t)微分可求得採用最高點(Peak Point)的時間𝑡 ∗ ，與t ∗ 時的累積採用者N(𝑡 ∗ )
及當期的採用者n(𝑡 ∗ )。
1
𝑝
ln(
)
𝑝+𝑞
𝑞

𝑡∗ =

N(𝑡 ∗ ) = m x

m(𝑡 ∗ ) = m x
1
2
(𝑝
+
𝑞)
4𝑞
1
𝑝
(2 - 2𝑞)
21
Bass的新產品擴散模型


整體而言，Bass模型中使用相當少的估計變數而能達成相當不錯的預測效果。Bass
模型及其修正模型常被用於預測產品擴散的途徑，尤其是在零售服務、工業技術、
教育、製藥以及耐久性產品等市場，皆能合理表達及解釋創新產品的擴散過程，並
且有效達到預測之目的。
但由於Bass模型是一個非常簡化的擴散模型，所以對一些重要的影響因素都採用簡
化的方法來處理，在1970年代之後，許多學者相繼提出質疑，並加以修正，其中以
Mahajan et al.(1985,1988,1990)提出最多的假設與模型應修正之處，詳細內容讀
者可自行參考他們的文獻內容。
22
創新擴散理論的應用



近年來創新擴散理論在資訊管理領域的應用，主要在「管理革新」、「
資訊系統與資訊科技導入」這兩個層面。在管理革新方面，例如Sharma
& Rai(2003)檢視資訊系統創新適應力以及領導者特質兩者間的關係，
發現職位上的權勢以及工作的任期正好與組織方面資訊系統創新的學習
呈負相關的線性關係，這與傳統關於組織行為的理論正好相反。
而在資訊系統與資訊科技導入的探討上，如Rajagopal(2002)藉由創新
擴散的觀點，並引用Kwon & Zmud(1987)的六階段模型，探討ERP系統作
為創新科技本身，其導入過程與影響企業建置ERP的因素。另外，
Bradford & Florin(2003)則以創新擴散理論以及資訊系統成功等理論
為研究基礎，發展出ERP系統建置成功的模型。
本節以Hu.Saunder & Gebelt在1997年的論文「資訊系統委外的擴散：
重新評估影響來源」(Diffusion of Information Systems
Outsourcing:A Reevaluation of Influence Sources)舉例說明如下。
23
研究問題


資訊系統委外在90年代曾經是個很熱門的概念，透過一些大公司的採用
，迅速在許多公司擴散，因而引起學者想要瞭解在這個擴散過程中，大
公司的採用是否會有帶頭作用。在1992年，Loh & Wenkatraman(LV92)
的研究使應用創新擴散數學模型來研究「資訊系統委外」(Information
System Outsourcing.ISO)影響來源，他們以60個公司的資料分析，認
為影響委外策略的擴散是來自「內部影響因素」與「柯達效應」(Kodak
Effect)所造成。柯達效應是指1989年柯達公司(Eastman Kodak)將大部
分的資訊系統、設施及人員成功委外後，可能帶動其他企業逐漸接受資
訊系統委外的管理方式，促成了委外市場加速成長。在LV92的研究中發
現在Kodak之前內部效果不顯著，但是Kodak之後則是以內部模仿效果主
導，也就是「柯達效應」是存在的。
Hu et al.(1997)的研究則認為之前LV92的研究在資料取得及研究方法
上都有些錯誤，因而進行進一步蒐集資料並建立四種擴散模型加以驗證
，認為柯達效應並不存在。其研究模式、方法與結論分別說明如下：
24
研究模式



該研究分為兩個階段。首先，他們用LV原來的資料，並且擴大蒐集了
175家公司的資料，分別建立了四種擴散模型來檢驗三個研究假說。
四種擴散模型(Diffusion Models)
若擴散速率是潛在採用者(Potential Adopters)和已採用者(Adopters)
之間差異的函數結果，其公式為：
𝑑𝑁(𝑡)
𝑑𝑡

= g(t)(m-N(t))
其中，N(t)為T=t的時間點，累積的採用者人數。
M為T=t的時間點，潛在採用者的人數。
g(t)為擴散係數(Coefficient Diffusion)，在不同的模型中，
其計算方式也不同。
25
研究模式

1.內部影響模型(Internal Influence Model)：代表創新擴散模型是藉由採用者
(N(t))與潛在採用者(m)之間相互溝通的方式，此時g(t)=qN(t)，q為內部擴散係數
。其公式為：
𝑚
N(t) = 𝑚−𝑚0
1+ 𝑚0 exp −𝑞𝑚𝑡

2.外部影響模型(External Influence Model)：此模型認為大眾傳播媒體(Mass
Media)是創新擴散的主要管道，此時g(t) = p，其公式為：
𝑑𝑁(𝑡)
𝑑𝑡

3.混合影響模型(Mixed Influence Model)：此模型認為創新擴散同時受到人際溝
通(Interpersonal)與大眾傳播媒體的影響，所以g(t) = p + qN(t)，混合模型的
公式為：
𝑑𝑁(𝑡)
𝑑𝑡

= p + q(N(t))(m-N(t))
4.馮‧貝塔郎菲影響模型(Von Bertalanffy “Influence Model)：作者認為前三
種引用Mahajan & Peterson (1985)的模型是非彈性的擴散模型(Inflexible
Diffusion Model)。Von Bertalanffy模型則為彈性擴散模型，其S型轉折點可動態
決定。其公式為：
𝑑𝑁(𝑡)
𝑑𝑡

= p(m-N(t))
=
𝑏
1−𝜃
𝑁 𝜃 (t)(𝑚1−𝜃 - 𝑁 1−𝜃 (t))
作者以上述四種擴散模型(Diffusion Model)及隨機的白雜訊模式(White Nose
26
Model)，驗證何種模型對資訊系統委外的擴散最具有最佳解釋力。
研究假定(Research Hypotheses)

該研究針對LV92的研究結論，設定了三個假說：
(1) 資訊系統委外的決策同時受到內部與外部因素的影響，所以混合
模型在用於描述其擴散過程時，應獲得較佳的結果。
(2) 有靈活的轉折點(Flexible Inflection Point)的模型至少和固
定轉折點(Fixed Inflection Point)模型一樣準確，因為沒有理
論或實證研究的基礎支持IS的委外擴散過程中轉折點是固定的。
(3) 內部影響模型適合解釋「前柯達時代」(Pre-Kodak Era)影響資
訊系統委外的主要因素，在「後柯達時代」(Post-Kodak Era)則
以混合影響模型為主。
27
研究方法


研究者以「資訊管理」 + 「委外」為關鍵字，查詢ABI-INFORM
Global、Newspaper Abstract、Periodical Abstract三個資料庫，
找到197家公司，可用的有1985年1月到1995年1月共175家有委外經驗
的企業樣本來檢驗擴散模型，檢驗方式有兩種：第一種是使用非線性
最小平方回歸法(NonLinear Least Squares Regression Method)來
估計四個模型的參數，第二種方法是以Davidson & MacKinnon(1981)
的J-test與P-test來檢查不同模式間的差異顯著性。
另外先以Mahajan et al.(1988)的白噪音模型(White Noise Model)
檢驗「擴散是隨機發生」的虛無假定，結果顯示為不接受，表示擴散
的過程不是隨機的過程(Random Process)，因此可以繼續進行假說檢
定，接著檢驗三個假說。
28
研究結論



研究結果顯示：
(1)混合模式確實比單純的內部或外部模式更佳；
(2)固定轉折點和彈性轉折點的兩種混合模式無法確定何者一定較佳；
(3)前柯達資料以Von Bertalanffy模型有較佳解釋力，後柯達時期和
前柯達時期看不出有明顯差異，因此假說應予拒絕。
該研究清楚的指出四種擴散模型中，以混合模型最能代表IS委外的擴
散過程，即「外部媒體」(External Media)，「賣方的壓力」
(Vendor Pressure)，與「內部溝通」(Internal Communications)這
三個因素，促使管理者決定採用資訊系統委外的方案。
此外，他們認為沒有足夠證據支持柯達效應的存在，因為柯達與IBM
等在1989年簽訂的委外案例，並沒有顯著的改變IS委外擴散的本質。
因為該研究使用的樣本數為175，比LV92的60個樣本數更具有代表性
，該研究也修正了LV92因模型與參數估計錯誤造成的外部效度不足的
問題。
29
Information technology innovation
diffusion: an information requirements
paradigm

Nigel Melville* & Ronald Ramirez† *Stephen M. Ross School of
Business, University of Michigan, Ann Arbor, MI 48109-1234, USA,
email: npmelv@umich.edu, and †Business School, University of
Colorado at Denver, PO Box 173364, Campus Box 165, Denver, CO
80217-3364, USA, email: ronald.ramirez@cudenver.edu

本研究主要是延伸許多探討創新擴散的文獻，提出以不同產業對資訊
需求影響IT的創新擴散，並以3個(程序複雜性、產業速度、供應鏈複
雜性)變數衡量資訊處理需求，以及2個(IT生產控制、線上供應鏈管
理)衡量資訊處理能力，發展研究架構。
本研究以上述研究架構，以美國的製造業彙總資料進行產業別分析，
發現不同產業別在不同IT有不同程度應用，透過比較選擇木製業及飲
料業進行研究架構的質性驗證。

30
Abstract




Information technology (IT) innovation research examines the
organizational and technological factors that determine IT adoption and
diffusion, including firm size and scope, technological competency and
expected benefits.
We extend the literature by focusing on information requirements as a
driver of IT innovation adoption and diffusion.
Our framework of IT innovation diffusion incorporates three industrylevel sources of information requirements: process complexity,
clock speed and supply chain complexity.
We apply the framework to US manufacturing industries using
aggregate data of internet-based innovations and qualitative
analysis of two industries: wood products and beverage
manufacturing.
31
Abstract


Results show systematic patterns supporting the basic thesis of the
information processing paradigm:
 higher IT innovation diffusion in industries with higher
information processing requirements;
 the salience of downstream industry structure in the adoption
of interorganizational systems;
 and the role of the location of information intensity in the
supply chain in determining IT adoption and diffusion.
Our study provides a new explanation for why certain industries were
early and deep adopters of internet-based innovations while others were
not: variation in information processing requirements
32
1
2
1-1
2-1
1-2
2-2
1-3
33
INTRODUCTION



The internet has changed the way firms conduct business, enabling new
value creating activities and supporting enhanced collaboration with
trading partners.
However, the use of the internet and its related impacts vary widely
across industries. In the personal computer (PC) manufacturing
industry, for example, internet-based technologies have had wideranging impacts throughout the value chain, transforming competitive
dynamics and generating new sources of value (Kraemer et al., 2000).
In contrast, diffusion of these technologies has been modest in other
industries such as textiles (Forza et al., 2000).
34
INTRODUCTION




According to the dominant diffusion of innovation paradigm, technology
characteristics, including trialability, observability, relative advantage,
complexity and compatibility might provide an explanation for such
differences (Rogers, 1983).
However, focusing on technology characteristics provides only a partial
explanation for industry variation because similar types of information
technologies (IT) are available to all firms and are readily implemented
by a burgeoning IT consultancy industry.
Another explanation provided by the dominant paradigm is technology
adopter characteristics, such as firm size and scope, technology
leadership, technology competency, readiness and expected benefits
(Rogers, 1983).
Indeed, empirical research shows the influence of such organizational
characteristics on innovation adoption (Mustonen-Ollila & Lyytinen,
2003).Given substantial intraindustry firm variation along these
dimensions (Chwelos et al., 2001; Macher et al., 2002; Zhu et al., 2004),
this explanation is also incomplete.
35
INTRODUCTION




Moving beyond the dominant paradigm, the competitive environment,
including competitive pressure and industry standards, may also play a
role in technology diffusion and adoption (Tornatzky & Fleischer, 1990).
In sum, although the dominant paradigm and extensions may provide
some insight into observed variation in internet adoption across
industries, it is not clear to what extent these or other factors may
explain such differences.
Reviewing the IT innovation literature, Fichman (2004) concludes that
most studies have used economic-rationalistic models, an
organizational level of analysis, and a logic that organizations
possessing a greater amount of the right stuff will display a greater
amount of innovation.
In this vein, studying industry-level phenomena holds great potential to
provide new insight into IT innovation adoption (King et al., 1994;
Crowston & Myers, 2004).
36
INTRODUCTION





We move beyond the dominant paradigm of IT innovation literature by
examining IT innovation from an information processing perspective.
The production process, supply chain and competitive environment vary
substantially across industries, creating differing information processing
needs.
According to the information processing view, enhancement of a firm’s
information processing capabilities is a viable response to such
information challenges and can be achieved through investments in
technology-based process improvements.
It would thus seem reasonable to posit that industry-level characteristics
are associated with variation in IT-based innovation adoption.
As a first step contributing to understanding in this area, we undertake
an analysis within the manufacturing sector. Our fundamental research
question is: How do information processing requirements
determine the adoption and diffusion of IT innovation across
industries?
37
INTRODUCTION






The structure of the paper is as follows.
We begin by describing the theoretical antecedents of the information
requirements paradigm of IT innovation diffusion, including diffusion
theory and the information processing view of the firm.
Next, we develop a conceptual framework linking information
processing requirements to information processing capabilities.
We then apply the framework as a tool to interpret empirical data on US
internet-based innovation adoption and alternative innovation profiles of
two contrasting industries: wood products and beverage manufacture.
In doing so, we show the validity of the framework and extend
knowledge of how information processing requirements shape IT
innovation adoption and diffusion.
The last section describes implications for research and practice and
identifies limitations of the study.
38
THEORETICAL BACKGROUND

Diffusion theory

Three stages of diffusion theory can be identified in the literature:
dominant, technology-organization-environment and emergent.
According to the dominant paradigm, the rate and pattern of the
adoption and diffusion of ideas, practices or objects through populations
of potential adopters is affected by the characteristics of both the
innovation and the adopter (Rogers, 1983). The multistage adoption
process is affected by the actions of ‘key adopters’, the resulting profile
being characterized by an S-shaped curve.
A drawback of the dominant paradigm, however, is its narrow focus
on only two innovation drivers, namely, technology and
organizational factors.


39
THEORETICAL BACKGROUND




Zhu et al. (2004) test this framework by analysing innovations
associated with the electronic pre-processing, negotiation, performance
and post processing of business transactions among firms via the
internet.
The authors find that internal factors – technology competence and
firm scope and size – and external factors – consumer readiness
and competitive pressure – are significant adoption drivers, providing
support for the inclusion of external factors in studies of IT
innovation diffusion.
We thus extend our focus to move beyond diffusion theory,
consistent with the suggestion of Cooper & Zmud (1992, p. 137)
that studies will be improved if they ‘adequately account for the
“fit” between the technology being examined and the work context
within which the technology is being introduced.’
We adopt the information processing view of the firm to shed light
on the connection between information processing requirements
across industries and the adoption and diffusion of IT innovations.
40
Information processing view






The information processing view provides an alternative rationale as to
why firms adopt IT innovations: to respond to information processing
requirements.
When changes in the environment occur, they give rise to uncertainty,
as established routines can no longer be used.
Firms face new information requirements as additional information must
now be collected, processed and distributed in order for the firm to
make decisions and complete its activities.
To continue to operate efficiently, a firm must align its information
processing capabilities with its new information requirements.
The firm responds to complexity by either increasing its information
processing capabilities or by reducing its information processing needs
(Galbraith, 1977).
IS scholars have used the information processing view to explore the
relationship between IS and information requirements.
41
Information processing view







Existing research shows that information processing capabilities
adopted in response to information processing requirements can
vary.
The information processing view in the IS context is also supported by
studies of organizational response to industry clock speed (Mendelson
& Pillai, 1998; 1999).
In highly dynamic business environments, firms must adopt new ways of
processing increased volumes of information, with IT being a major
enabler of such processing.
Results support the information processing view of the firm in that
higher clock speed environments are associated with greater use of IT
by firms.
In sum, the information processing view is useful in the context of ITbased innovation as it provides an alternative perspective of why firms
adopt such innovations.
Internal and environmental uncertainty, whether based in production
methods, supply chains or the larger competitive landscape, gives rise
to new information requirements – a key driver of firm adoption of
information technologies.
We now develop a conceptual framework of IT-based innovation based
on the information requirements paradigm.
42
CONCEPTUAL FRAMEWORK






The conceptual framework is developed to shed light on our primary
research question, which asks how information processing requirements
shape the adoption and diffusion of IT across industries.
The basic thesis is that within an industry, common means of
production, supply chains, and a shared competitive landscape provide
common information requirements for firms within that industry, leading
to incentives to adopt IT innovations.
As an example, firms in industries with short product life cycles and
rapid product innovation are characterized by a high degree of
uncertainty, leading to substantial information processing requirements.
Firms may turn to various forms of IT as a means of providing the
information processing capacity necessary for rapid product
development and market introduction.
In contrast, all else being equal, we might expect industries with longer
product life cycles to be subject to less stringent information processing
requirements, resulting in lower IT adoption.
We build on prior research by developing a conceptual framework that
includes information processing requirements (process complexity,
clock speed and supply chain complexity) and information processing
capabilities (production systems and interorganizational systems)
(Galbraith, 1977; Tushman & Nadler, 1978; Bensaou & Venkatraman,
1995; Mendelson & Pillai, 1998).
43
CONCEPTUAL FRAMEWORK







Information processing requirements arise from the gap between
information required and information available, i.e. uncertainty.
Three sources of uncertainty are (1) task complexity and subunit task
interdependence; (2) task environment; and (3) interunit task
interdependence (Tushman & Nadler, 1978).
The framework is intended to represent interdependencies between the
three task characteristics: uncertainty, fit and information processing
requirements.
We adapt the three sources of uncertainty used in the original
framework as follows.
To match our approach of analysing IT innovations across industries,
we include process complexity to reflect the focus of modern firms on
business processes and provide a contemporary construct
encompassing the original task construct.
The second dimension – clock speed –encompasses the rate of
environmental change that was included in the original framework and is
adopted from prior IS research (Mendelson & Pillai, 1998).
The third dimension is supply chain complexity, which extends the idea
of external interdependence of tasks to the contemporary business
environment of complex, global supply chains.
44
CONCEPTUAL FRAMEWORK



Regarding information processing capabilities, we focus on salient
categories of IS used in the manufacturing environment: IT-based
production control and supply chain management systems.
(As our focus is not on other information processing mechanisms, such
as meetings and site visits, we do not include them in our framework.)
We model the five measures as formative indicators of their
respective constructs, as the direction of causality flows from measure
to construct and the measures do not necessarily covary (Jarvis et al.,
2003) (Figure 1).
45
Information processing requirements

Process complexity

Process complexity includes a variety of factors associated with the
production process that give rise to uncertainty (Table 1).
First is product variety, which denotes the number of parts, models,
brands, etc. High product variety raises complexity and uncertainty due
to multiple interrelated demand signals (Prasad, 1998).
Related to the type of product being produced, the second process
complexity factor – production method – includes job shop, flow,
continuous processing and assembly, each showing alternative levels of
process complexity (Woodward, 1994).
The third process complexity dimension is information intensity of the
production process, including the volume of information required for
quality control steps, physical tolerances, etc.



1-1
46
47
Information processing requirements
1-2

Clock speed

Clock speed denotes the speed of industry evolution, including the rate
at which new products, processes and organizational structures are
introduced (Fine, 2000).
Resulting from endogenous industry factors such as competition and
technological change, specific measures include the duration of product
life cycles and the rate of decline in the prices of input materials
(Mendelson & Pillai, 1998).
Firms in high-clock speed industries face extremely short design and
production horizons, as well as severe pressure on inventory (Fine,
1998).
Many well-known cases of internet-based innovation appear in highclock speed industries such as microprocessors (Phan, 2003) and data
routers (Kraemer & Dedrick, 2002).
In an industry such as PC manufacturing with rapidly declining prices,
controlling inventory costs is a key to success, and internet-based
technologies are used to manage supply chains using build-to-order and
online sales (Kraemer et al., 2000).




48
Information processing requirements

Supply chain complexity

Supply chain complexity denotes the degree of coordination required
across firm boundaries.
This includes the informational and product span across physical and
logical spaces as well as the nature of interactions between upstream
suppliers and downstream buyers.
For example, the apparel industry is characterized by a global end-toend supply chain of design, production and distribution (Teng &
Jaramillo, 2005).
In contrast, wood products tend to be processed near the source of raw
materials (forests), and then trucked or shipped to plants that use
processed timber as an input (e.g. home builders).
Such a difference in the geographic dispersion of information and
production associated with supply chain processes presents alternative
needs for the use of interorganizational IS, including both traditional EDI
as well as web-based interorganizational IS.
Also salient is the institutional structure of the demand channel, which, if
highly concentrated, may afford bargaining power to pressure upstream
49
firms to adopt IT innovations.





1-3
Information processing capabilities

IT-based production control

Computers and networks are an integral part of the manufacturing
production process.
Mainframe, minicomputer and PC-based systems enable computeraided design and computer aided manufacturing (CAD/CAM).
System functionality includes analytic applications used in forecasting,
scheduling and other operations management functions as well as
digital schematic representation and manipulation.
An auto manufacturing plant, for example, may use programmable logic
controllers and distributed control systems to collect sensory data,
check it against control parameters, and send feedback signals to
production machine tools.
A PC-based production system may also use analytic software for
forecasting and scheduling.
Given variation in demand dynamics and process complexity across
industries, we expect that different forms of production control are used
across industries.





2-1
50
Information processing capabilities
2-2

Supply chain management

EDI is an industry standard method of communicating electronic
documents, such as purchase orders and invoices, via proprietary value
added networks.
In recent years, the internet and its open standards has emerged as a
platform of choice for communication and exchange across the supply
chain, subsuming EDI functionality and adding an array of real-time
collaboration and exchange functions (Johnson & Whang, 2002).
Internet-based exchange and collaboration also support global supply
chains such as those found in the electronics manufacturing industry.
Extranets, for example, enable trading partners to access portions of
internal data, thereby increasing information processing capabilities
(Fowler et al., 2000).
Real-time customer inventory data are communicated back to the
supplier so that accurate decisions can be made regarding supplierrelated manufacturing schedules and finished goods stock levels. There
is wide variation in such internet-based innovation across industries,
with empirical evidence suggesting an informational basis for such
variation (Cagliano et al., 2003).



51







ANALYSIS
As reviewed above, prior research has not focused on examining how
information processing requirements shape the adoption and diffusion
of IT innovations across industries.
Moreover, the majority of studies analysing internet-based innovations
have been conducted in high-tech industries, leaving uncertainty as to
what is happening in other industries.
Based on these considerations, we have chosen to apply our framework
using descriptive industry data analysis and focused industry
studies.We conduct descriptive data analysis of 21 North American
Industry Classification System (NAICS) industries and qualitative
analysis of two industries: wood products and beverage manufacturing.
While the former grounds our perspective in aggregate empirical
observation, the latter provides a rich tableau of the variation in
information processing requirements and information processing
capabilities across two varying industries.
Cross-industry analysis is important for uncovering phenomena that
may be localized to a particular industry and have not been exposed in
prior industry research (Chiasson & Davidson, 2005).
Indeed, our selection of industries provides such an opportunity as a
majority of industry-related research examines manufacturing, hightechnology, and finance, real estate, and insurance industries.
In sum, the advantage of our analysis approach is that it provides
richness and scope, while the key limitation is the lack of formal
52
statistical analysis.
Aggregate US industry data







We examine industry-level data to provide insight into how internetbased innovation adoption and diffusion vary across industries due to
variation in information processing requirements.
Although few data sources and little prior research has been carried out,
there is a database collected by the US Census Bureau at the NAICS
industry level. The data do not allow for empirical analysis at the firm
level as published data comprise tabulations of NAICS industries.
Further, they do not enable formal statistical analysis, as there are only
21 three-digit NAICS industries.
However, they do provide a unique lens into the use of internet business
processes across roughly 30,000 US manufacturing firms in 21
industries for the year 2000.
Five technologies that underlie internet-based innovation are intranets,
the internet, local area networks (LAN), EDI and extranets.
Based on our developed conceptual framework, we might expect a
greater percentage of firms in industries characterized by a high degree
of process complexity, rapid clock speed, and high supply chain
complexity to use these technologies.
To examine whether this might be the case, we tally the number of
instances in which a given industry appears in the top five or bottom five
in terms of percentage of plants using these five technologies (Table 2).
53
54
Aggregate US industry data




Another perspective on internet-based innovation is to move beyond
technologies and examine functionality of innovations.
Figure 2 displays five basic processes – email with trading partners,
internet buying, EDI buying, internet selling, EDI selling – according to
the percentage of using plants (lowest industry, median, highest
industry). Internet use and EDI use are included to provide a baseline
comparison.
The data indicate that three industries account for the highest
percentage of plants using these processes: computer and electronics,
transportation, and printing and related. The first two are consistent with
data on technology usage presented in Table 2.
Industries with the lowest percentage of adopting plants include
petroleum and coal, wood, non-metallic metals, textile mills, and printing
and related. Once again, we see clear consistency with these industries
and those that adopt technologies slowly.
55
Figure 2. Percentage of manufacturing plants adopting e-business innovations
by industry: rank ordered. Source: Tehan (2003). Note: Industry names provided
for highest and lowest categories, i.e. those with the highest percentage of plants
in each particular category as well as the lowest. Median percentage is indicated
by the square.
56
Aggregate US industry data



Although preliminary and subject to firm-level empirical testing,
exploratory analysis of industry-level data appear to be consistent with
our thesis that information processing requirements are an important
factor in the extent to which industries adopt internet-based innovation.
To examine this hypothesis further, we examine one industry at the low
end of the usage scale – wood products – and one in the middle –
beverage manufacture.
We do not examine highly innovative industries as these have been
studied extensively in prior research.
57
Industry application: wood products and
beverage manufacturing





Data for the industry analyses were gathered through a variety of
sources. First, we used research databases to identify pertinent industry
reports and published research involving any aspect of computerization
or supply chain management. Second, we examined company reports,
such as 10K (年度業績報告)documents, of firms within the two selected
industries. Third, we examined newspaper articles and government data
on internet business practices.
The resulting corpus of material provided a substantial amount of raw
data enabling analysis of each construct in the conceptual framework in
great detail.
We apply the conceptual framework within two industries to show its
applicability and to inform our fundamental research question. Several
case studies exist of firms in industries with high degrees of all three
dimensions of information processing requirements. Not surprisingly, in
these industries it is found that internet-based innovation is abundant.
In contrast, we sought low-tech industries with moderate and low
degrees of information processing requirements. To provide the
necessary variation in information processing requirements, we sought
variation in both manufacturing processes (process complexity) and the
interfaces between suppliers and customers (supply chain complexity).
We chose wood products (NAICS 321) as it is consistently ranked in the
bottom of internet-based innovations in Census Bureau data and
beverage manufacturing (NAICS 3121) as it is consistently in the middle
58
range.
Industry application: wood products and
beverage manufacturing

Wood product manufacturing

Wood product manufacturers produce wood products such as lumber,
plywood, veneers, wood containers, wood flooring, wood trusses,
manufactured homes, and prefabricated wood buildings.
On the demand side, buyers of industry output use processed lumber
for furniture, home construction, cabinets, flooring, and the like.
In the primary segment, lumber, plywood and particle board are sourced
to sales distribution channels including wholesalers, distributors,
brokers and retail stores.
On the supply side, upstream is the logging process, which includes
forest harvesting, sorting and handling.
A simplified schematic of the wood product manufacturing supply chain
is shown in Figure 3a.




59
60
1-1






Information processing requirements:
process complexity
The multistage process of creating wood products from raw timber
begins with the scanning of a raw roundwood piece of lumber for the
existence of unwanted foreign material that could adversely affect the
production process (Araman et al., 1992).
After the scan, the log is debarked, graded for quality, and an
individualized production plan is developed to maximize the log’s
processed value.
A band hedger creates boards from the log according to plan and the
boards are trimmed and then kilned to remove moisture.
After drying, the boards are planed, graded and stacked for delivery to
customers (Table 3).
Other than the size and grade of the finished board, there is a low
degree of product variety.
Moreover, there are relatively few production stages: scan, prep, cut
and trim, kiln, finish.
61
1-1




Information processing requirements:
process complexity
From an information perspective, when a piece of raw roundwood
lumber is scanned before edging, information is collected on defects,
the amount of clear wood, and potential board cutting plans.
This information is used to determine standard products that can be
generated from each log, to grade the lumber according to industry
standards, and to determine the best cutting plan for the maximization
of finished product value (Araman et al., 1992).
While the cutting of each log is relatively straightforward, uncertainty
regarding individual log shape, size and quality requires the
development of production processes that take into account
nonstandardized inputs.
There is a specialized front-end process that necessitates the creation
of dynamic production plans for each log entering the lumber mill. In
sum, process complexity is low compared with other industries.
62
1-2







Information processing requirements:
clock speed
There is a relatively low rate of change in the types of wood products
demanded by downstream buyers as well as in the upstream wood
procurement inputs.
Moreover, processes and organizational structures are fairly stable
(Vlosky & Smith, 2003).
Wood products are intermediate goods used in the creation of homes,
furniture and other end products.
Consumer preference changes are reflected in the manufacturing of
these finished goods rather than the output of lumber mills.
In addition, while technology such as geographic IS is sometimes used
to manage timber inventory and yields, lumber inputs (logs) themselves
do not have embedded technologies.
As such, there is extremely low technological innovation and pressure in
the upstream supply of logs as well as in the downstream demand for
lumber.
In sum, the stability of lumber preferences and technology advancement
suggests a relatively low clock speed relative to other industries.
63
Information processing requirements:
1-2
clock speed






Regarding competitive pressure, lumber plants vary in size depending
upon which of the two major categories of wood they manufacture:
softwood or hardwood.
While the softwood lumber industry is concentrated with larger firms
operating at a larger scale, the hardwood lumber industry is relatively
fragmented, with companies employing less than 20 workers (Vlosky &
Smith, 2003).
Regardless, all sectors in the lumber industry are facing increased
environmental pressures.
Manufacturers face increased competition from global firms, especially
those from the Canadian softwoods industry.
Lumber companies have responded by reducing overall production and
closing downs mills (Cumbo et al., 2003). Environmental regulations
and pressures to achieve sustainable forest development also provide
an impetus for change, although it is unclear to what extent this actually
impacts the production plans of industry participants.
Overall then, the low rate of change in inputs and products and modest
degree of competitive pressure create a relatively low-clock speed
environment in wood product manufacturing.
64
65
66
1-3







Information processing requirements:
supply chain complexity
The supply chain for wood product manufacturers is more informationintense toward the customer or downstream end (Figure 3).
Upstream, timber mills are often located near forests, so that the
interface is geographically close.
Moreover, from an information perspective, trees take years to mature,
meaning that the only source of variation is in the final quality of the
logs, which can be monitored for minimal impact on production
schedules of timber mills.
There is a greater degree of complexity on the demand side of the
supply chain than on the supply side.
Wood product manufacturers, for example, sell their products to a host
of customers including wholesalers, distributors, retail chains, brokers
and builders.
This results in a complex web of information exchange as
manufacturers conduct multiple transactions with multiple downstream
partners (Vlosky et al., 2002).
In addition, with regard to their large retail customers, lumber
manufacturers face increasing pressure to seamlessly capture and
transmit information related to lumber product transactions.
67
1-3



Information processing requirements:
supply chain complexity
This is especially true for companies such as Home Depot that
increasingly look to informate and automate the procurement function in
order to improve the efficiencies. It is not clear, however, to what extent
buyer concentration may provide them with the power to dictate their
information processing needs to their suppliers, i.e. wood product
manufacturers.
The upstream portion of the wood products supply chain is important
but because of the low variation of raw timber logs entering the
manufacturing process, the information requirements are not as intense
as that at the downstream end.
The small relative number of raw timber suppliers also reduces the
overall information needs of the downstream supply chain.
68
2-1




Information processing capabilities:
IT-based production control
Certain wood product manufacturers use analytic software extensively
to control production within required tolerances. Analytic software is
used for scanning and optimization decisions at the front end of lumber
production.
Edger optimization and automated grading systems help manufacturers
produce higher quality products on a more consistent basis (Bowe et al.,
2002).
However, although efficiency enhancing, analytic systems are not
uniformly adopted within the lumber industry.
Only 10% of hardwood manufacturing firms use some sort of advanced
scanning and optimizing technology due to the scarcity of resources for
these relatively small firms (Bowe et al., 2002), consistent with
aggregate industry-level data presented earlier.
69
2-2





Information processing capabilities:
supply chain management
Wood product manufacturing firms have made minimal investments in
computer networking and interorganizational systems.
In a recent survey of primary wood products manufacturers, Vlosky &
Westbrook (2002) find that although a majority of respondents have a
web site, not a single one used it for electronic transactions with trading
partners. Preferred methods include telephone, mail and fax.
Moreover, when internet applications are used, they are used primarily
to support financial transactions, such as invoices, order
acknowledgements, advanced shipping notices and purchase orders.
These are narrow applications that are not integrated with other
business functions such as production operations (Vlosky & Smith,
2003).
While computers are being used to enter business transaction
information, networking-based applications such as email and EDI are
used only sparingly (Vlosky & Westbrook, 2002).
70
Beverage manufacturing



Beverage manufacturers produce non-alcoholic beverages, fermented
alcoholic beverages, and distilled alcoholic beverages (Figure 3b).
The major beverage categories are soft drink, bottled water, ice, beer,
wine and distilled spirits.
As a result of the relatively small size of the distilled spirits market, we
focus our discussion on soft drinks, beer and wine products.
71
1-1





Information processing requirements:
process complexity
The beverage production process comprises six stages: raw materials
management, product creation, bottling, packaging, warehousing and
distribution.
Although various final products are produced, from beer to cola, the
primary process difference between beverages occurs in the product
creation stage.
To emphasize, other than the product creation stage, beverage
manufacturing shares five similar production stages.
Raw ingredients are ordered from multiple suppliers and managed in
order to minimize inventory cost and protect against stock outs.
Beverage products are bottled and put into a variety of package types,
with final goods being warehoused and distributed to a multitude of
wholesalers and retailers.
72
1-1






Information processing requirements:
process complexity
Regarding information, the production process is relatively complex.
At the raw materials stage, for example, beer manufacturers must
analyse the quality of barley, including data on moisture content,
nitrogen level and grain weight (Kourtis & Arvanitoyannis, 2001).
In product creation, knowledge and information is required for machine
set-up, line processing design, performance measurement and dynamic
process improvement to minimize raw material and product loss (Koss,
2003).
Production processes must also incorporate regulatory requirements to
ensure strict standards are met for each product being created.
For example, grape wine production must incorporate regulatory
information from the Code of Federal Regulations, Title 27, which
requires producers to limit the addition of dry sugar to no more than
20% weight, of water to no more than 10% weight of water, and the total
alcohol content derived by fermentation of no more than 13% by
volume. (Section 4.21, The Standards of Identity, Title 27: Alcohol,
Tobacco, and Firearms, http://ecfr.gpoaccess.gov/)
Similar regulatory requirements exist in bottled water, beer and
73
numerous other beverage products.
1-1




Information processing requirements:
process complexity
Information is also important for warehousing and distribution, as sales
order data is processed and interorganizational collaboration with
trading partners occurs to minimize product handling and the amount of
finished goods inventory maintained in the beverage manufacturing
warehouses (Esper & Williams, 2003).
However, despite the amount of information required in beverage
manufacturing, the production of beverages is relatively routine.
Product recipes are well known and after a production line is set up
according to an established design, the creation of a beverage follows a
standardized procedure using standard inputs.
In summary, given the safety, quality and degradability considerations in
this industry, as well as the need for precise tolerances with regard to
product and process recipes, production complexity is clearly higher
than that of wood products.
74
1-2





Information processing requirements:
clock speed
Similar to wood products, the rate of change in the beverage industry is
relatively moderate.
Product life cycles are typically quite long, although new brands are
introduced regularly (Cioletti, 2003a).
Also, input materials – raw ingredients and packaging – as well as
fundamental bottling processes do not change rapidly. However, in the
realm of competition, beverage production differs substantially from
wood product manufacturing.
The beverage industry is much more concentrated and contains much
larger firms on average.
Coca Cola and Pepsi accounted for 77% of the 2002–2003 carbonated
soft drink sales volume in the USA (Hemphill, 2004). (In wine
manufacturing, the top 30 companies account for 90% of wine market
sales. while the top three wine companies – E & J Gallo, Constellation
Brands, and The Wine Group – account for over 60% of all sales, this
sub sector is highly fragmented because of recent growth in the
premium wine segment (Wine Business Monthly, 2004).)
75
1-2







Information processing requirements:
clock speed
Such high levels of concentration can reduce overall competition within
an industry as firms may find it easier to coordinate their activities and
price above market levels (Bain, 1951).
While this may result in less of an incentive to innovate, recent research
suggests an inverted U-shaped relationship between innovation and
competition (Aghion et al., 2005).
A second dimension of clock speed that is salient in this industry is
regulation.
A recent issue within the beverage industry that is raising information
needs is the 2002 Bioterroism Act.
This regulatory act requires all food and beverage firms to establish the
capability to track products from raw ingredients to the end consumer.
Beverage firms are applying their resources and capabilities towards
capturing information along the beverage supply chain in order to meet
this government imposed information requirement (Folwell & Volanti,
2003; Higgins, 2003a).
Overall then, clock speed is moderate in this industry.
76
1-3





Information processing requirements:
supply chain complexity
The beverage manufacturing supply chain is information intensive
relative to wood products due to upstream and downstream information
requirements.
Manufacturers that produce soft drink products, for example, must work
with multiple suppliers to coordinate shipment of syrups for colas, root
beer, fruit drinks and a host of other carbonated beverages.
Information must also be shared with suppliers for the procurement and
delivery of water, bottles and CO2 gas.
On the downstream side, soft drink manufacturers collaborate with
distributors, vendors and fountain operation customers.
Interorganizational data sharing is a necessity for a seamless logistics
process and adds to the overall information requirements of the
beverage supply chain.
Also, given the importance of minimizing finished goods inventory, the
information shared in the downstream portion of the supply chain is
important to beverage manufacturers for product forecasting and
production planning processes. Unforeseen delays or miscues in the
supply chain can have severely detrimental fiscal impacts as inventory
buffers are no longer available to guard against such issues.
77
1-3



Information processing requirements:
supply chain complexity
As an example, inadequate sharing of customer information and
production planning led the Coors Corporation to experience product
shortfalls during recent high demand seasons such as Christmas and
New Year holidays, resulting in lost sales and upset customers
(Moozakis, 2001).
Another aspect of the supply chain is the concentration in certain
downstream customer segments such as grocery stores. This creates
buyer power which may act to force beverage producers to adopt IT
innovations to better manage downstream physical and logical
interfaces.
In sum, supply chain complexity is higher than that of wood product
manufacturing.
78
2-1





Information processing capabilities:
IT-based production control
In some beverage product markets, like premium wine and distilled
spirits, minimal usage of IT in production is driven primarily by image.
For these types of products, the identity of the brand is more important
to sales than the unattractive image of computer-driven stainless steel
manufacturing (Elliott, 1998).
In contrast, analytic technology such as checkweighers, automated
metal detectors and quality control software has become important in
soft drink and other beverage manufacturing (Higgins, 2002).
For example, Coriolis meters are a widely used control technology
which incorporates digital processing to control the flow of syrups, juices
and other liquid-based beverage ingredients (Higgins, 2003b).
Beyond automation and control technology, customer-facing IT
solutions also affect the beverage production process. Specifically, soft
drink sales personnel use hand-held computers to capture and transmit
real-time major customer demand data which is interpreted by
production planning systems and translated into more efficient
production scheduling (Krell, 2004).
79
2-2







Information processing capabilities:
supply chain management
A widely used technology in beverage manufacturing supply chain and
front end activities is Enterprise Resource Planning (ERP) systems,
used for supply chain management and partner collaboration (Hayes,
2001).
Anheuser Busch uses an SAP module for inventory management of
parts for its plant equipment (Fryer, 1998).
Workers use the system to obtain part descriptions, quantity available
and other pertinent data. Coca Cola also uses an ERP system to
integrate with its supply chain partners for best practice information
sharing, resource sharing and to capitalize on economies of scale
(Violino, 1999).
Networking technology has become an important tool for meeting
today’s information requirements.
Along the supply chain, for example, Coca Cola is undertaking a
network-based project to enhance data sharing and synchronization
with its upstream suppliers (Foley & Kontzer, 2004).
In the downstream portion of the supply chain, the Coors Corporation
has turned to networking technology to integrate distributors with its
production planning system.
Through enhanced information sharing via an extranet, Coors hopes to
eliminate future production shortfalls during high demand seasons
80
(Moozakis, 2001).
DISCUSSION





Our analysis has shown systematic patterns that reinforce and refine
the information processing perspective applied to IT innovations in
general and internet-based innovations in particular.
We find a positive association between information processing
requirements and internetbased technology usage.
Further examination shows that an industry’s supply chain structure is a
driver of information processing requirements and that the location of
information requirements is an important factor in the diffusion of IT
innovations, especially internet technologies.
Finally, results indicate that regulation can influence internet-based
innovation across industries.
We now discuss these findings in detail.
81





First, aggregate descriptive statistics spanning 21 US manufacturing
industries show that the use of technologies associated with internetbased innovation tends to be higher in industries with greater
information processing requirements.
The PC manufacturing industry, which is high in all three dimensions of
information processing requirements, is in the top five of usage in four of
five internet-based innovation technologies and leads in terms of email
with trading partners.
In contrast, the wood products industry is in the lowest category in all
technology usage measures and several innovation dimensions.
This finding extends prior understanding of e-commerce adoption
antecedents that identified technology competence, firm scope and size,
consumer readiness, and competitive pressure as significant
determinants (Zhu et al., 2004).
By expanding the focus to underlying information processing
requirements faced by all firms in an industry, our framework adds a
layer of understanding to existing knowledge.
82





Aggregate descriptive analysis also indicated anomalies, such as the
printing industry having the highest percentage of firms adopting
internet buying, but the lowest percentage of firms adopting internet
selling.
This finding shows that although we have focused on three dimensions
of information processing requirements, other factors also come into
play.
Example factors identified in earlier research include market power and
the adoption of third-party business-to-business (B2B) portals in textile
industries (Cho, 2006) and the potential loss of social capital and the
adoption of B2B electronic marketplaces in the beef industry (Driedonks
et al., 2005).
While our analysis contributes to the overall understanding of industry
factors driving internet-based technology adoption, examining the
relative importance and potential interrelationships among information
processing requirements and other drivers is an interesting avenue of
future research.
The second finding is that supply chain structure is an important
dimension of information processing requirements, driving enhanced
use of internet-based innovations to control the flow of inbound
materials and outbound products.
83







Downstream buyers of beverage products include large, national firms
such as retail grocery chains.
Given the amount of transactions with a few of these large buyers, the
beverage industry is sensitive to their needs, and partners to ensure
systems integration across the supply chain and smooth flow of product.
In contrast to the customers of beverage manufacturing, buyers of wood
products are heterogeneous, encompassing wholesalers, distributors,
retail brokers and builders.
Each buyer has its own particular needs, but none has sufficient
purchasing power to drive widespread adoption of sophisticated
demand-side electronic integration.
This finding sheds new light on the adoption drivers of
interorganizational systems such as EDI (Fichman, 2004).
Indeed, when combined with earlier evidence (Wigand et al., 2005), our
results provide an explanation for minimal adoption of EDI technologies
in certain industries: supplier and buyer characteristics.
Further research is needed to identify additional aspects of supply
chains that drive IT innovation.
84







The third finding is that the location of information requirements makes
a difference with respect to the diffusion of IT innovations, especially
internet technologies.
For lumber manufacturers, the most intense information requirement
occurs at the front end of the production process due to the natural
variation in raw timber entering the manufacturing facility.
For beverage manufacturers, on the other hand, multiple raw inputs are
required to produce alcoholic and non-alcoholic beverages.
This imposes a different information requirement vs. lumber
manufacturing, as the use of multiple inputs requires information flows
to determine and track individual input quality, quantity and the timing of
when the input is applied within the beverage production process.
The nature of procurement, materials and upstream supply industries is
thus also an important dimension of information processing
requirements that drive IT innovation diffusion.
The fourth finding is that regulation may drive or constrain internetbased innovation across industries.
A key difference in information processing requirements between wood
and beverage manufacturers arises from a new regulatory act that
requires the tracking of beverage products throughout the supply chain.
85






This, in turn, increases the information requirement of beverage
manufacturers as they must collect and track data related to its
suppliers and manufacturing centres.
While our fourth finding is unique to industry-level research on internet
technology adoption, it is consistent with existing research examining
the adoption of other technologies such as those in the US petroleum
refining industry (Chen, 2005).
Overall, our findings indicate that variation in IT adoption and diffusion
across industries appears to be due at least in part to industry variation
in information processing requirements.
This adds to our knowledge of innovation adoption and diffusion
phenomena.
Limiting the domain of inquiry to factors in the dominant paradigm and
extensions would not have yielded such insights.
As an example, differences between firms in an industry with medium
information processing requirements and those in a low information
processing requirements might confound analyses based on firm or
technology characteristics, which may lead to insignificant or biased
results.
86
CONCLUSION







Prior research shows that and technology characteristics affect the
extent to which firms adopt technological innovations. Other factors also
play a role.
We built on prior research to develop a conceptual framework for an
information processing requirements paradigm of IT innovation
diffusion.
To examine the validity of the developed framework, we examined a
database of internet-based innovation in the US manufacturing sector.
We supplemented this descriptive analysis of aggregate statistics by
examining two industries in detail that vary along the three dimensions
of information processing requirements: wood product manufacturing
and beverage manufacturing.
Results of our study have several implications for research.
First, the information requirements paradigm is a useful lens through
which to view IT innovation adoption and diffusion and thereby extend
the dominant paradigm.
By emphasizing the notion of fit between informational needs and
processing capabilities afforded by IT, we developed a new perspective
on industry phenomena and innovation adoption.
87








This is consistent with Fichman’s (2004) call to move beyond the
dominant paradigm in IS innovation research.
In future research, we plan to build on the foundation developed herein
by constructing and validating measures using firm-level data from the
US Census Bureau and formally testing the framework developed
herein.
(Data on more than 35 internet-based innovations in roughly 30 000 US
manufacturing plants have been collected by the US Census Bureau; a
proposal to the US Bureau of the Census, Center for Economic Studies
to use these data has been approved.)
Second, using qualitative industry analysis, we have identified
phenomena pertinent to each construct of the framework and salient to
particular industry settings.
This aids future researchers by demonstrating how higher-level
constructs might be operationalized.
Third, by examining low-tech, slower-moving industries, we have
showed the need to expand the breadth of research focus beyond
technological innovators.
Examining why innovation is not happening can be as informative as
examining why it is.
Fourth, as we focused on specific measures of internet-based
innovation, such as extranets and EDI, there is a need to examine
additional types of IT, such as vendor-managed inventory and
88
collaborative design systems.







Finally, a natural extension is to examine two dimensions of difference –
traditional firm-level characteristics as well as industry-level phenomena
– within the same analysis.
This would answer the important question of the extent to which each is
important and in what contexts.
An important implication for managers is that although the information
processing requirements facing firms within a given industry may show
structural uniformity, the response of firms does not.
We attempted herein to illustrate the variation in structural patterns of
responses across industries.
However, there is not always a fit between the two, and in some
industries, information capabilities may lag requirements.
In such industries, there is great potential to achieve improved
productivity and efficiency by leveraging the knowledge of firms in more
advanced industries.
This is a fruitful area of future research as well as the starting point for
managerial application of the developed framework.
89







Finally, despite our attempts to limit threats to validity, our study
suffers from several potential limitations.
First, we did not carry out any formal statistical tests due to the
aggregate nature of the Census data and lack of other available
data sources at present.
Other factors might be at work, such as the size of firms, the degree
to which the industry is concentrated, the path-dependent nature of
innovation, and the like.
Moreover, standard measures of reliability and validity could not be
calculated because of the limited nature of the data. Another
potential limitation is that we were not able to incorporate withinindustry variation in both information requirements and capabilities.
Our data are cross-sectional, so we were not able to study the
dynamic nature of industries over time.
Specifically, information processing requirements may not be stable
over time, and their evolution would also likely affect the evolution of
technology adoption and diffusion.
(We thank an anonymous reviewer for underscoring this point.)
90







Our descriptive approach to applying the framework to US industry data
must therefore be interpreted with caution.
Also, the model does not incorporate all the factors driving a firm’s IT
application in production or in its supply chain.
Application of the framework to the wood product and beverage
manufacturing industries was based on secondary data, which may
suffer from inaccuracies.
Also, it is difficult to estimate the degree to which findings based on
these two industries may generalize to others.
Given our examination of several other industries in selecting these two,
as well as careful analysis of prior research of high-tech, high-clock
speed industries, it is our preliminary belief that findings will generalize.
At the least, systematic patterns in quantitative and qualitative data
showed herein point to the need for more detailed examination of these
issues.
Despite these limitations, then, the structural consistency in our findings
across the two methods provides a sound conceptual point of departure
for future research.
91
伍、結論與建議
92